Prompt detection of incipient faults and accurate monitoring of engine deterioration are key aspects for ensuring safe operations and planning a timely maintenance. Modern computing capabilities allow for more and more complex tools for engine monitoring and diagnostics. Nevertheless, an underlying physics-based approach is often preferable, because not only the “what” but also the “why” can be identified, providing an effective decision support tool to the service engineer.
In this work, a physics-based adaptive model is used to evaluate performance deltas and correct the data to reference conditions (gas turbine load and ambient conditions), while a data-driven correlation algorithm identifies the most likely matches within a fault signatures database. Possible faults are ordered from the highest correlation in the decision support system and the most likely fault can be selected based on the number of occurrences and the associated correlation. Gradual engine degradation can also be monitored by displaying performance deltas trends during time.
The diagnostics tool was tested on a validated performance model of a single-shaft industrial gas turbine and subsequently on experimental data. This paper presents the diagnostics system structure, the model adaptation scheme, and the results obtained from simulated and real fault data. Accurate fault isolation and severity identification were achieved in all cases, demonstrating the tool capability for decision support system.